International Journal of Agricultural Science Research Vol. 2(5), pp. 131-143, May 2013 Available online at http://www.academeresearchjournals.org/journal/ijasr

ISSN 2327-3321 ©2013 Academe Research Journals

Full Length Research Paper

Biometric and morphometric studies of Perna viridis and Perna indica along the southwest coast of India: A

statistical approach

Jayalakshmy, K.V.1*, Maheswari Nair1, Dileepkumar, R.1 and Vijayan, M.2

1National Institute of Oceanography, Regional Centre, Kochi-682018, Kerala, India.

2Central Pollution Control Board, Delhi, India.

Accepted 8 February, 2013

The growth indices of the mussels, Perna viridis and Perna indica collected from the south west coast of India were examined using statistical methods. The Principal Component Analysis (PCA) and Exploratory Factor Analysis (EFA) separated 2 morphometric gradient groups (spat and adult), indicating the different phenotypic plasticity between them. The factor scores classified Perna indica as a unimodal, positively skewed leptokurtic population and Perna viridis as a unimodal, negatively skewed leptokurtic population. The allometry was not static, but simple and ontogenetic since their population was continuously varying. The present study shows that factor analysis is better than principal component analysis for delineating the morphometric characteristics of living organisms. Key words: Mussels, morphometric, biometric, skewed, principal component, factor analysis, kurtosis.

INTRODUCTION Mussels are bivalve mollusks, attached to rocks or any other hard substratum. The common mussels Perna indica and Perna viridis predominantly inhabit along the littoral regions of oceans. Growth patterns of mollusks are important to evaluate their production potential and energy flows through populations (Gaspar et al., 2002). Perna viridis is a common amphiboreal species in India and its growth is influenced by environmental and climatic conditions. Of the two species, the green mussel, Perna viridis, is widely distributed both on the east and west coast of India along the inter-tidal zones, while Perna indica has restricted distribution. The Perna indica shell grows up to 12 cm long and 5 cm thick, while that of Perna viridis grows up to 23 cm long and 7 cm thick. The European counterpart of the green mussel is Mytilus edulis which are found in low intertidal estuaries. Perna viridis is of commercial importance because of rapid growth and abundance. They are indicators of pollution by heavy metal, organic chlorides and petroleum hydrocarbons. It is generally tolerant up to salinity 80 PSU, but also survives at reduced salinities. Perna viridis begins its life as a juvenile with a green and blue green

shell that develops brown patches as an adult. This is distinguished from all others by having 30 diploid chromosomes instead of 28. The growth characteristics of mussels are important to understand the influence of climatic conditions on their morphometry. Temperature is the main factor influencing their growth and other factors include the habitat like littoral, sub-littoral and estuarine regions (Kulakovskii and Lezin, 2002). In the present study, we examine the statistical tools on the morphological features in the two mussels, which are very common in south west coast of India. MATERIALS AND METHODS Thirty samples of the two mussels, Perna indica and Perna viridis were collected from two stations (Vizhinjam and Calicut) of south west coast of India (Figure 1) and

*Corresponding author. E-mail: kvjayaparam@yahoo.co.in. Tel: +91-084-2390814.

Jayalakshmy et al. 132

Figure 1. Study region showing station locations.

the samples were immediately taken for morphometric measurements. This involves the measurement of shell length (LT), height (HT), width (WI), volume (VL), shell weight (WT), wet weight (WW), total weight (TW) and dry weight (DW) of the two mussel species. LT, HT and WI were measured in the laboratory using the vernier calipers accurate to 0.01 cm. WT, WW, TW and DW were taken using an electrical digital balance accurate to 0.001 gm. In addition, three different ratios were defined: WI/LT,

HT/LT and WI/HT for the two sets of samples to determine whether a simple allometry exists between the size and shape measurements or it is a complex allometry. Data analysis All data subjected to statistical tests were first checked for normality using chi-square test of goodness of fit

(Sokal and Rholf, 1995) applying correction for chi square test and using the degrees of freedom as n-p-1 where n is the number of classes and p is the number of parameters estimated for fitting normal distribution (Anderson, 2006) and for homogeneity of variances using Bartlett’s test (Snedecor and Cochran, 1967; Jayalakshmy, 1998). When these conditions were satisfied, parametric tests were used in subsequent analysis otherwise non-parametric analogues were used. Pearson correlation coefficient test was used to analyse the correlations between the various characteristics of mussels. Before applying the normality test, the data were classified into frequency distribution using Sturge’s formula for number of classes (Gupta, 2007) to minimize the error due to grouping and all the distribution parameters including the test of symmetry and kurtosis of the distribution are calculated (Jayalakshmy, 1998; Gupta, 2007) to justify the type of population from which the sampling was done.

Ordination of samples by PCA and EFA were applied to compare the two species based on the linear combination of the morphometric measures and then to determine a linear combination to discriminate between the species based on these measures which reflect the similarity of their morphometric measures. In this study, both PCA and EFA are applied and the groups of measures obtained are compared between the species to delineate the type of distribution of the population from which the samples are drawn. Factor loadings of factors 1 and 2 are plotted for measures as well as for mussels to compare the sets of measures with that obtained from the PC analysis. As an additional application of EFA, distributions of factor scores for mussels were explored (Goodall, 1954) to designate the type of population from which the species samples were collected. Statistical techniques such as 2 Way ANOVA and student’s t test can identify the differences between their characteristics based on allometry. Canonical discriminant analysis was applied to discriminate between the two populations of mussels based on the morphometric measures which are not significantly correlated (Snedecor and Cochran, 1967; Jayalakshmy, 1998). RESULTS Perna indica collected from Vizhinjam ranged in length from 3.05 to 6.65 cm (Figure 2a to h, Table 1) while Perna viridis collected from Calicut measured 2.43 to 10.6 cm in length (Figure 3a to h, Table 1). Shell height of P. indica varied between 1.87 and 3.20 cm whereas for P. viridis it was 1.44 to 4.47 cm. Shell width varied over a narrow range (1.05 - 1.95 cm) for P. indica while for P. viridis, it was higher (0.87 - 3.3 cm). Shell volume of P. indica was only 1/4th (1.5 to 5 ml) of that of P. viridis (0.52 to 22 ml). Mussel wise variation for volume was double in the case of P. viridis (C.V.% = 92.98) compared to P. indica. Shell weight for P. indica ranged over 1.25 to

Int. J. Agric. Sci. Res. 133 7.07 gm whereas it was 1.55 to 33.1 gm for P. viridis. Mussel wise variability of shell weight for P. viridis was nearly 3 times that of P. indica with shell wet weight ranging over 1.28-4.26 gm. In the case of P. viridis, the range for shell wet weight was 1.18 to 22.8 gm, nearly 5 times that of P. indica, being nearly 3 times bigger than P. indica. Dry shell weight varied in the range of 0.11 to 1.65 gm for P. indica and 0.03 to 3.62 gm for the other species. Normal distribution (Jayalakshmy, 1998) fitted for the two species showed that the observed measurements (Table 1) adhere to Normal probability density function (P>0.05) for all measures except wet weight and total weight in the case of P. indica and except height and wet weight in the case of P. viridis (Table 1) (P<0.05). Invariably, for all shape measures (length, height, width and volume), P. viridis showed higher significance and higher deviation from normality when compared to P. indica, whereas for size measures, it was vice-versa (Figures 2 and 3). Blay (1989) has studied the length distribution of 5 populations of the fresh water bivalve Aspatharia sinuate in Nigeria.

Q-Mode factor analysis (Gooddall, 1954) when applied for P. indica after row normalization and varimax rotation to simple structure (Kaiser 1958), delineated only one significant factor with eigen value, λ, 29.58 explaining about 49.79% variability forming the differential factor group and this factor contains all the mussels from 1 to 19 and the factor 2 with eigen value of 0.2933, explaining about 49.79% of the total variance, constituted by the individuals from 20 to 30 (Figure 4a). The factor 1 contained individuals of lower values for all parameters, whereas factor 2 contained mussels with higher values for all parameters, thus (Harman, 1967) grouping the mussels into spat- small mussels and adult -large mussels. But spat mussels are observed to explain the same amount of variations observed in the morphometric characteristics as that by the adult mussels collected from Vizhinjam implying that P. indica spat animals are ecologically equally efficient as their adults.

PCA applied has also divided the 30 mussels into two PCs with eigen values of 29.58 and 0.293 respectively and each explaining 49.79% of variance. PC1 and PC2 plotted for P. indica has divided the 30 mussels into two groups. One with 19 mussels which are big shaped and big sized (adults) mussels and the second with 11 mussels which are small shaped and small sized (spat) ones (Figure 4b). EFA has also grouped the mussels exactly in the same manner as obtained from PCA. But the EFA has given a clear cut demarcation between mussels based on their size and shape for both the species and this uniqueness has been materialized only due to varimax rotation to simple structure. Also all measures are observed to be significantly correlated with factor 1 and factor 2 loadings (r>0.901, P<0.001) except VL (r=0.891, P<0.01) and HT (r=0.736, P<0.01) which are moderately correlated and it is clearly depicted in the plot of factor loadings also (Figure 4a to d).

Jayalakshmy et al. 134

Fre

qu

en

cy

Fre

qu

en

cy

Fre

qu

en

cy

Fre

qu

en

cy

F

req

uen

cy

Fre

qu

en

cy

Fre

qu

en

cy

Fre

qu

en

cy

Figure 2. Normal distribution fit to the morphometric measures. (a) Shell volume, (b) Shell width, (c) Shell height, (d) Shell length, (e) Shell dry weight, (f) Shell total weight, (g) Shell wet weight, (h) Shell weight for Perna indica.

Int. J. Agric. Sci. Res. 135

Table 1. Distribution of shell measurements of Perna indica and Perna viridis.

Parameter Mean Std CV% Inter quartile range β1 Test 1 β 2 Test 2 Calculated value of χ2+

Distribution of shell measurements of Perna indica

Length 4.979 0.8901 17.88 0.712 0.1443 0.3226 2.0928* 2.342* 0.1944

Height 2.685 0.337 12.55 0.2358 -0.0333 -0.0744 2.4105 2.695* 2.3333

Width 1.6077 0.2365 14.71 0.1961 -0.1429 -0.3196 2.1524 2.4065* 2.5774

Volume 2.6333 1.1442 43.45 0.9046 0.7008 1.5677 2.2174 2.4792* 0.1917

Weight 4.1473 1.6176 39.07 1.3095 0.3604 0.8058 1.8371 2.054* 2.6705

Wet wt. 2.8013 1.0443 37.28 0.9558 0.4507 1.0077 1.5794 1.7658 15.3095*

Total wt. 6.9337 2.6138 37.70 2.1256 0.4901 1.0959 1.6858 1.8848 11.379*

Dry wt. 0.5693 0.3256 57.19 0.154 1.3243 2.9613* 5.2334 5.8511* 4.0657

Distribution of shell measurements of Perna viridis

Length 5.7053 2.2623 39.65 1.9756 1.194 2.6699* 3.426 3.8304* 4.0536

Height 2.7380 0.7919 28.92 0.6972 0.4990 1.1177 1.8088 2.0223* 6.75*

Width 1.8153 0.6483 35.71 0.5988 0.5069 1.1334 1.9195 2.1461* 3.9821

Volume 6.0167 5.5948 92.99 2.8907 1.2279 2.7457* 3.6712 4.1045* 2.6389

Weight 8.446 8.5398 101.11 9.5439 0.4542 1.0157 1.8918 2.115* 2.5635

Wet wt. 7.3613 5.4538 74.09 4.0459 1.0763 2.4066* 3.4212 3.825* 7.6727*

Total wt. 15.6413 13.7404 87.85 7.7978 1.1652 2.6054* 3.4694 3.8789* 3.8259

+, Tabled value of χ2(2) = 5.99, *, Calculated value of corrected frequency (expected frequency < 5 were clubbed together with adjacent frequencies) χ2(2) > 5.99, P<0.05.

R-mode factor analysis applied on P. indica resulted in three factors of which only factor 1 is statistically significant (λ >1) explaining a total of 99.61% of total variance with highest eigen value of 7.76, explaining about 44.26% of the variability in the mussel distribution (Figure 4c). All loadings of the first factor were negative with moderate loadings for LT, HT, WI, WT, WW and TW, and almost approximately equal loadings for first three measures. Second factor has all positive factor loadings with almost same factor loading for all measures except for DW for which this factor has highest loading with eigen value, 0.1410 and explains about 33.07% of the variability among mussels. This analysis further stresses the fact

that to describe P. indica, the morphometric measures including DW are highly essential and additional measures are also required explaining this mussel because the total variance explained by factors 1 and 2 is less than 100 by a value of 22.67% which is not negligible.

Principal component analysis carried out for measures of P. indica resulted in 4 PCs and explained about 97.5% of total variance (Figure 4d). The first PC explained the maximum of the total variation among the samples (88.3%). All of the first component loadings were strongly positive for all morphometric variables except ratio measures, but with almost equal value with respect to all measures. It indicates that PC1 is a

measure of shell size due to the high proportion of the explained total variance. Further, it can be concluded that mussel size accounted for most of the variance in the data. Approximately, 10.7% of total variance was explained by PC2. Loadings on the PC2 were all positive except that of WI and WW (which are very low values) and DW which has the maximum negative loading on PC2. PC3 explained about 5.9% of total variance in the data. The high positive loading of PC3 was for HT. PC4 had all the loadings negative except that of DW which was the maximum positive for this factor and for HT and TW which were the second maximum and very low values respectively. PC analysis plotted for grouping of measures for P.

Jayalakshmy et al. 136

Fre

qu

en

cy

Fre

qu

en

cy

Fre

qu

en

cy

Fre

qu

en

cy

Fre

qu

en

cy

Fre

qu

en

cy

F

req

uen

cy

Fre

qu

en

cy

Figure 3. Normal distribution fit to the morphometric measures. (a) Shell volume, (b) Shell width, (c) Shell height, (d) Shell length, (e) Shell dry weight, (f) Shell total weight, (g) Shell wet weight, (h) Shell weight for Perna viridis.

Int. J. Agric. Sci. Res. 137

Figure 4. (a) Principal component loadings for PC1 and PC2 of Perna indica, (b) Factor loadings for factors 1 and 2 of Perna indica, (c) Principal component loadings for PC1 and PC2 of morphometric measures of Perna indica, (d) Factor loadings for factors 1 and 2 of morphometric measures of Perna indica.

indica, delineated three separate clusters keeping shape measures separated from size measures and all weight measures together with DW as a different entity. Further PCA clustered the ratios in a highly associated manner while the shape measures were grouped in a dispersed pattern, but more associated than the size measures.

Q-mode analysis applied on P. viridis collected from Calicut, delineated factor 1 with eigen value (λ), 28.56, explaining about 48.72% and constituted by mussels with lower values for all parameters and factor 2 with eigen value, 0.955 explaining about 48.15% of the variability and constituted by mussels with higher values for all parameters, both together forming the differential factor groups (Figure 5a). This analysis has thus grouped P. viridis also based on shell size determined by various weight measurements and shell shape determined by HT and LT, as small sized- small shaped and large sized - big shaped mussels. The plot of PC1 and PC2 obtained

from PCA applied for mussels of P. viridis has divided the total of 30 individuals into two as small mussels constituting the PC1 and large mussels with larger shape constituting the PC2 (Figure 5b). Factor analysis has also grouped the mussels exactly in the same manner as obtained from PC analysis but with a clear cut characterization for the same reason mentioned earlier (Figure 5a).

Factor analysis by R-mode for the morphometric measures of P. viridis resulted in 1 significant factor explaining (λ > 1) 42.75% of total variance (Figure 5c). The first factor explained only less than half of the total variation among the data measures. All of the first factor loadings are strongly negative and approximately equal with respect to WT, TW, WW and VL. Also, the second factor has positive loadings for all the measures and with almost high positive loadings for HT, LT and WI for which it has low negative loadings on factor 1. 42.45% of the

Jayalakshmy et al. 138

Figure 5. (a) Principal component loadings for PC1 and PC2 of Perna viridis, (b) Factor loadings for factors 1 and 2 of animals of Perna viridis, (c) Principal component loadings for PC1 and PC2 of morphometric measures of Perna viridis, (d) Factor loadings for factors 1 and 2 of morphometric measures of Perna viridis.

total variance was explained by the second factor also. The third factor explained only a nominal amount (14.42%) of the total variance. It also has very low positive loadings on all variables except DW. Hence it indicates that the first two factors constitute measures of shell size with respect to high proportion of the total variance and form the differential factor groups including all the weight measurements except dry weight. Hence, it could be concluded that the P. viridis is almost completely defined by the morphometric measures premeditated at present, which constitute size factor, shape factor and carbon factor. It further emphasizes that no additional characteristics are required to identify this species because almost all of the variability could be extracted from these three factors unlike P. indica.

For P. indica, Q-mode Factor 2 contained measures which are highly consistent where as factor 1 is that of measures which highly inconsistent (Table 1) are. In the

case of P. viridis also the two groups of morphometric measures are those of higher scale of variability forming factor group 1 and lower scale of variability (Table 1) forming factor group 2. Also, all measures are observed to be significantly correlated with factor 1 loadings and factor 2 loadings (r>0.934, P<0.001) except DW (r=0.709, P<0.01) and HT (r=0.876, P<0.01) which are only moderately correlated with the factor loadings.

Based on PC analysis carried out for P. viridis, 4 PCs are delineated with a total of 97.2% of the total variance in the measures being explained (Figure 5d). PC1 has all positive loadings with approximately same value for all measures. PC2 has high negative loading on WI and very low negative loading on WW and TW whereas PC3 has moderately high negative loading on DW. PC1 explains about 74.6% of the total variance while PC2 explains only 12.9% of the total variance, concluding that mussel size accounted for most of the variance in the data. PCA for

Int. J. Agric. Sci. Res. 139

Table 2. Significance of skewness and kurtosis of the factor score distribution of Perna indica and Perna viridis.

Perna indica Perna viridis

Factor score β1 Test1 β2 Test2 No. of modes Remark β1 Test1 β2 Test2 No. of modes Remark

Factor 1 1.1402 2.5496* 3.7914 0.8848 1 PSLD -1.8869 -4.2193** 5.7698 3.0967** 1 NSLD

Factor 2 1.4859 3.3225** 5.7146 3.0350** 1 PSLD -1.7314 -3.8705** 5.6163 2.9251** 1 NSLD

Factor 3 -0.5269 -1.1782 2.7294 -0.3025 2 NSPD 0.4253 0.9509 3.9528 1.0653 1 PSLD

Factor 4 0.4444 0.9936 2.9048 -0.1064 1 PSPD 0.08245 0.1833 1.7193 -1.4319 2 PSPD

Factor 5 -0.3023 -0.6759 2.8290 -0.1912 1 NSPD -1.2882 -2.8806** 4.8325 2.0488* 1 NSLD

Factor 6 -0.6309 -1.4107 3.0858 0.0959 1 NSLD -0.8784 -1.9642 4.2044 1.3466 1 NSLD

Factor 7 -0.1725 -0.3856 3.0609 0.0681 1 NSLD -0.2555 -0.5713 2.4811 -0.5802 1 NSPD

Factor 8 2.2175 4.9584** 9.5807 7.3574** 1 PSLD 0.4236 0.9471 2.6154 -0.4300 1 PSPD

PSLD - positively skewed leptokurtic distribution; NSLD - negatively skewed leptokurtic distribution. PSPD - positively skewed platykurtic distribution; NSPD - negatively skewed platykurtic distribution.

P. viridis showed ratios to be highly associated than the shape measures and the size measures. Factor loadings of factors 1 and 2 when plotted for P. viridis showed a similar pattern with that obtained from PCA, but were more closely associated, while the shape and size were more virtually separated. These disparities in these clusterings were less for P. indica. This may be because the mussel-wise variation was more for the latter species compared to the former (Table 1). Studies have compared the ratios of several morphometric variables in mussels (Dermott and Munawar, 1993; Pathy and Mackie, 1993), D. polymorpha, D. bugensis (Andrusov, 1897) and Mytilopsis leucophaeata (Conrad, 1831). In general, the disadvantage of this type of analysis is that ratios are not constant within a group that shows allometry at substantial plasticity (Reyment et al., 1984). Simple allometry occurs if the ratio between the specific growth rates of two different characters is constant (Blackstone, 1987). In this study, the ratios, WI/LT, HT/LT and WI/HT are almost constants having less variability (Table 1) indicating simple allometric relationships between

the size measures. Factor score distribution when subjected to shape and symmetry showed that factors 1 and 2 for both the species are significantly skewed, positively for P. indica and negatively for P. viridis and amount of Kurtosis in the population appears to be non trivial for factors 2 and 8 for P. indica and factors 1, 2 and 5 for P. viridis (Table 2). For both the species factor, score distribution is unimodal for all factors except factors 3 and 4 for P. indica and P. viridis respectively. This indicates that the samples of both the species have been generated from a continuously varying population but not from distinct and discontinuous mussel populations. Two way ANOVA has been applied to compare the two species of mussels for the significance of the difference between the measured characteristics. All morphometric variables except that of HT and WI/LT differed significantly between P. indica and P. viridis at 1% level (P < 0.01). Between mussels also, the measured variables differed significantly (P < 0.05) except that of WI/HT (P > 0.10). Student’s t test applied to compare the measures also resulted in the same conclusion. Two way ANOVA as well as t

test applied for comparing between factor loadings and between mussels showed significant difference (P< 0.01) between the 2 species for factors 3 and 4. Difference between the mussels of the 2 species was also highly significant for factors 1, 2 and 3. Factor loadings of the morphometric measures compared between the 2 species showed high difference only for HT but at a lower level of significance (P < 0.10). Mussel wise difference was also highly significant between species for all the measures except DW (P < 0.05). The morphometric measures could discriminate between the two species with 89% cross validation (Figure 6). DISCUSSION The commonly used parameters to study growth kinetics of bivalves are the shell length and height (Franz, 1993; Zainudin and Tsuchiya, 2007). However, these data may not always reflect the mass of an organism. Factors such as reproduction, population density and habitat are found to influence the tissue growth. The morphometric variations for P. viridis are greater

Jayalakshmy et al. 140

Figure 6. Discriminant analysis for classification of Perna indica and Perna viridis.

than P. indica, as the latter are larger in shape. WT and VL varied widely for P. viridis, whereas DW and VL varied for P. indica. The smaller size of P. indica may be attributed to the different environmental conditions (at Vizhinjam and Calicut) and the presence of predators at Vizhinjam (mussels are consumed by fishes like leatherjackets, crabs and starfish). However, mussels growing in spat fall region will be easily subjected to fouling (by barnacles, tunicates, bryozans, algae, oysters and amphipods), when their growth is inhibited.

The influence of biological environment is stronger since there is a habitat difference in the proportion of LT to TW (Black, 1977; Lewis and Bowman, 1975). Similarly, the intra specific variation is found between the two populations of C. karachiensis from the Gulf of Oman and the Arabian Gulf and has been attributed to the difference in the habitats. The intra specific variations between the two species in WI/LT, WI/HT and HT/LT ratios could be due to the difference in the habitat and size between the species. Alternatively, a difference in the morphometric

character (Emam, 1994) explains the difference in the habitats of the two regions.

The distribution of measures for the two species generally adhered to the probable normal density function except WW and TW for P. indica and HT and WW for P. viridis. For all shape measures (LT, HT, WI and VL), P. viridis invariably showed greater significance and higher deviation from normality as compared to P. indica, whereas it was vice-versa for size measures (Figures 2 and 3). Both PCA and EFA are data reduction techniques. But the limitations for PCA are its rigidity of dissimilarity measure and weak distance preservation. The first problem is solved by Gower (1966) extending the PCA to Principal Coordinate Analysis (PCOA) or classical scaling. In this method, the set of variables in higher dimensions are reduced to linear combinations of these variables. PCA is used when the variables are highly correlated to reduce the number of variables which account for most of the variance observed.

Factor analysis was used to identify the underlying constructs in data that cannot be directly measured. This method was used to identify factor structure without imposing any preconceived structure on the outcome (Angel and Fasham, 1974; Child, 1990). EFA provides a distribution pattern for the population from which the sample is collected. It is observed that Q-Mode analysis in EFA (Harman, 1967) grouped the mussels into 2 distinct classes of small and large sizes, but there was uniform variation in their morphometric characteristics. R mode analysis for P. indica reveals that other morphometric measures are highly essential since the total variance explained by factors 1 and 2 are <100. However, this species is completely defined by the morphometric measures of size, shape and carbon factor, emphasizing the better prediction of this procedure.

The PCA divided the 30 mussels (P. indica) into two groups (PC1, PC2) of big (19 mussels) and small (11 mussels) size. Factor analysis also grouped the mussels exactly in the same manner. But EFA distinguished mussels based on their size and shape through varimax rotation to simple structure. There was also significant correlation between factor 1 and factor 2 loadings (P<0.001) except for volume and height, which were moderately correlated. The total variance of morphometric data for zebra mussel (Dreissena polymorpha) was explained by the shell size alone (Trichkova et al., 2008), whereas both shape and size contributed to the total variance in the present study. Taxonomical studies on quagga mussel (Dreissena bugensis) did not yield any nucleotide difference, suggesting that the profundal form of the quagga mussel is a phenotype and not a separate species (Claxton et al., 1998). In contrast, the second and third PCs of the morphometric variables including length, width, height and weight separated the epilimnetic and profundal forms of the quagga mussel irrespective of the depths (Mackie,

Int. J. Sci. Res. 141 1991). This is due to the fact that D. bugensis shows plasticity in shell morphology with respect to depth, whereas D. polymorpha does not.

All morphometric variables except HT and WI/LT differed significantly between P. indica and P. viridis. Between mussels also, the measured variables differed significantly except for WI/HT.

Student t test also gave the same result. The growth rate of mussels may vary intra or inter-specifically depending on a number of factors including tides, seasons (Sutherland, 1970), food availability (Parry, 1977), maturation (Kay and Magruder, 1977) and eventually lead to different proportions and longevities (Branch, 1981).

The results obtained by EFA and PCA were in broad agreement with each other except in grouping of mussels.

FA providing distribution patterns for the measures are useful for modeling studies. Unlike cluster analysis, FA does not impose a hierarchical structure on the data and the allometry between the measures is simple and ontogenetic, since the mussel population vary based on the factor scores (Cock, 1966). PCA and cluster analysis do not consider rotation to simple structure. The varimax rotation to simple structure (Goodall, 1954) has the advantage that the factor loading matrix (Morrison, 1978) determined from an interspecific coefficient matrix would be that each species would have high loadings on only a few factors (usually one) and its loadings on the other factors would be near zero. This makes the division of the variables into groups of associated variables simpler than if no rotation had been carried out. FA is better than PCA in this study for delineating the morphometric measures which uniquely defined the mussel population. This study further strengthens the fact that separate allometric relationships between length, height and width for spat larval and adult stages are essential to predict the corresponding size of the mussel.

Taxonomic characteristics have higher potential, while quantitative analysis provides morphological characters in an objective manner. Multivariate analysis can identify independent characters for separating morphological forms (Dodson and Lee, 2006).

The growth and shape of shells are influenced by abiotic (environmental) and biotic (physical) factors (Miguel et al., 2002). Abiotic factors include geographical location (Beakelma and Meehan, 1985), depth (Claxton et al., 1998), shore level (Franz, 1993), tides (Dame, 1972), currents (Furman et al., 1999), turbulence (Bailey and Green, 1988), waves (Akester and Martel, 2000), type of bottom (Claxton et al., 1998), sediment texture (Newell and Hadeu, 1982) and burrowing behavior (Eagar, 1978).

In this study, the mussels taken from two different environments were found to be varying by size and shape and hence, other factors are also to be included to further identify their morphometric behavior.

Jayalakshmy et al. 142 CONCLUSION Two mussels, P. indica and P. viridis collected from Southwest coast of India were subjected to morphometric studies using statistical techniques on shape and size growth parameters. PCA and EFA separated 2 morphometric gradient groups justifying the different phenotypic plasticity between the spat and adults of the two mussels. The population distribution of P. indica was unimodal, positively skewed and leptokurtic, whereas that of P. viridis was also unimodal and leptokurtic but negatively skewed. The growth allometry was simple and ontogenetic but not static since the population continuously varied. Further this study revealed that EFA is a better tool than PCA for delineating the morphometric characteristics of living organisms. ACKNOWLEDGMENTS The authors express their sincere thanks to the Director, National Institute of Oceanography Dona Paula Goa and the Scientist in charge, NIO Regional Centre Kochi, for providing the facilities used to carry out this work. REFERENCES Akester RJ Martel AL(2000). Shell shape, dysodont tooth

morphology, and hinge-ligament thickness in the bay mussel Mytilus trossulus correlate with wave exposure. Can. J. of Zoo., 78: 240–253.

Anderson MJ (2006). Distance based tests for homogeneity of multivariate dispersions. Biometrics, 62: 245-253.

Andrusov NI (1897). Fossil and living Dreissenidae of Eurasia. Proc. St Petersburg Social Nat., 25: 285-286.

Angel MV, Fasham MJR (1974). Further factor analyses of the plankton data. J. Mar. Biol. Asso. U K, 54: 879-894.

Bailey RC, Green RH (1988). Within-basin variation in the shell morphology and growth rate of a fresh water mussel. Can. J. Zoo., 66: 1704-1708.

Black R (1977) Population regulation in the intertidal limpet Patelloida alticostata (Angas, 1865). Oecologia, 30: 9-22.

Blackstone NW (1987). Specific growth rates of parts in a hermit crab Pagurus longicarpus A reductionist approach to the study of allometry. J. of Zoo., 211: 531-545.

Blay J (1989). Morphometric length–weight relationships and length distributions of five populations of the freshwater bivalve Aspatharia sinuate (Unionacea, Mutelidae) in Nigeria. Malacologia, 30: 365–372.

Branch GM (1981). The biology of limpets: physical factors, energy flow and ecological interactions. Oceano. Mar. Biol. Annual Rev., 19: 235-380.

Child D (1990). The essentials of factor analysis, second edition, Cassel Educational limited, London.

Claxton WT, Wilson AB, Mackie GL, Boulding EG (1998).

A genetic and morphometric comparison of shallow- and deep-water populations of the introduced dreissenid bivalve Dreissena bugensis. Can. J. Zoo., 76: 1269–1276.

Cock AG (1966). Genetical aspects of metrical growth and form in animals. Quartanery Rev. Biol., 41: 131-190.

Dame RF (1972). Comparison of various allometric relationships in intertidal and subtidal American oysters. Fish. Bull. (US), 70: 1121-1126.

Dermott R, Munawar M (1993), Invasion of Lake Erie offshore sediments by Dreissena, and its ecological implications. Can. J. Fish Aqua. Sci., 50: 2298-2304.

Dodson I, Lee CE (2006). Recommendations for taxonomic submissions to hydrobiologia. Hydrobiology, 556: 1-5.

Eagar RMC (1978). Shape and function of the shell: a comparison of some living and fossil bivalve mollusks. Biol. Rev., 53: 169–210.

Emam WM (1994). Morphometric studies on the limpet cellana karachiensis, (Mollusca, Gastropoda) from the Gulf of Oman and Arabian Gulf. Indian J. Mar. Sci., 23: 82-85.

Franz DR (1993) Allometry of shell and body weight in relation to shore level in the intertidal bivalve Geukensia demissa (Bivalvia: Mytilidae). J. Exp. Mar. Biol. Ecol., 174: 193-207.

Franz DR, Tanacredi JT (1993). Variability in growth and age structure among populations of ribbed mussels Geukensia demissa (Dillwyn)(Bivalvia; Mytilidae), in Jamaica Bay, New York (Gatewaya NRA). The Veliger, 36: 220-227.

Fuiman LA, Gage JD, Lamont PA (1999). Shell morphometry of the deep sea protobranch bivalve Ledella pustulosa in the Rockall Trough, north-east Atlantic. J. Mar. Biol. Asso. UK, 79: 661–671.

Gaspar MB, Santos MN, Vasconcelos P (2002). Shell morphometric relationships of the most common bivalve species (Mollusca: Bivalvia) of the Algarve coast (southern Portugal). Hydrobiol., 477: 73–80.

Goodall DW (1954). Objective methods for the classification of vegetation. III. An essay in the use of Factor analysis. Austra. J. Bot., 2: 304-324

Gower JC (1966). Some distance properties of latent root and vector methods used in multivariate analysis. Biometrica, 53: 325-338.

Gupta KL (2007). Statistical Methods, 6th Edition. IBH

publishing company, Harman HH (1967). Modern factor analysis. University of

Chicago press, Chicago Jayalakshmy KV (1998). Biometric studies on trophic

level relations in the Indian Ocean. Ph.D. Thesis. Cochin University of Science and Technology, Cochin.

Kaiser HF (1958). The varimax criterion for analytic rotation in factor analysis. Psychometrika, 29: 115-129.

Kay EA, Magruder W (1977). The biology of opihi

(Department of planning and economic development,

Hawai. Kulakovskii EE, Lezin PA (2002). Filtration activity of the

blue mussel, Mytilus edulis (Mytilida, Mytilidae), under influence of the starfish, Asterias rubens (Forcipulata, Asteriidae). Zoologichesky Zhurnal, 81: 292–297.

Larson RJ, Marx ML (1981). An Introduction to Mathematical statistics and its applications. Prentice Hall Inc, New Jersey.

Lewis JR, Bowman RS (1975). Local habitat-induced variations in the population dynamics of Patella vulgate L. J. Exp. Mar. Biol. Ecol., 17: 165-203.

Mackie GL (1991). Biology of the exotic zebra mussel, Dreissena polyrpha, in relation to native bivalves and its potential impact in Lake St. Clair. Hydrobiology, 219: 251-268.

Miguel BG, Miguel NS, Paulo V, Carlos CM (2002). Shell morphometric relationship of the most common bivalves species (Mollusck bivalvia) of the Algarve coast, Southern Portugal. Hydrobiol., 477: 73-80.

Morrison DF (1978). Multivariate statistical Methods. Mc. Graw-hill International Book Company.

Newell CR, Hidu H (1982). The effects of sediment type on growth rate and shell allometry in the soft shelled clam Mya arenaria (L.). J. Exp. Mar. Biol. Ecol., 65: 285–295.

Parry GD (1977). Life history strategies of five species of inter tidal limpet. Ph.D. Thesis, Mellbourne University, Australia.

Int. J. Agric. Sci. Res. 143 Pathy DA, Mackie GL (1993). Comparative shell

morphology of Dreissena polymorpha Mytilopsis leucophaeata and the Quagga mussel (Bivalvia: Dreissenidae) in North America. Can. J. of Zoo. 71: 1012-1023.

Reyment RA, Blackith RE, Campell NA (1984). Multivariate Morphometrics, 2

nd edition. Academic

Press, London. Snedecor GW, Cochran WG (1967). Statistical Methods.

Oxford and IBH Publishing Co, New York. Sokal RR, Rholf FJ (1995). Biometry, 3rd edn. Freeman,

New York. Sutherland JP (1970). Dynamics of high and low

populations of the limpet “Acmea Scabra” (Gould). Ecol. Mono. 40: 169-188.

Trichkova T, Kozuharov D, Hubenov Z, Botev I, Zivkov M, Cheshmedjiev S (2008). Characteristics of Zeebra mussel (Dreissena polymorpha) populations in infested reservoirs north west Bulgaria. J. of Nat. Hist., 42: 619-631.

Zainudin M, Tsuchiya M (2007). Growth of bivalves used on allometric relationship and a time series of length-frequency data. J. Sust. Sci. Manage., 2: 1-1.